A bootstrap-ultralean patch

A bootstrap-ultralean patch

[H.J. Lu]

Hi,

I made this patch to allow

# make bootstrap-ultralean

which only builds the C compiler for stage 1. It saves both CPU
time and disk space.

Thanks.


-- 
H.J. Lu (hjl@gnu.org)
--
Thu Jan 15 17:00:42 1998  H.J. Lu  (hjl@gnu.org)

	* Makefile.in (bootstrap-ultralean): New target.

Index: gcc/Makefile.in
===================================================================
RCS file: /home/work/cvs/gnu/egcs/gcc/Makefile.in,v
retrieving revision 1.1.1.17
diff -u -r1.1.1.17 Makefile.in
--- gcc/Makefile.in	1998/01/14 03:09:43	1.1.1.17
+++ gcc/Makefile.in	1998/01/16 01:48:18
@@ -2662,7 +2669,7 @@
 	  $(LANG_DIFF_EXCLUDES) \
 	  gcc-$(oldversion) gcc-$(version) > diffs
 
-bootstrap bootstrap-lean: force
+bootstrap bootstrap-lean bootstrap-ultralean: force
 # Only build the C compiler for stage1, because that is the only one that
 # we can guarantee will build with the native compiler, and also it is the
 # only thing useful for building stage2.
@@ -2673,9 +2680,12 @@
 # To prevent `make install' from compiling alloca.o and then relinking cc1
 # because alloca.o is newer, we permit these recursive makes to compile
 # alloca.o.  Then cc1 is newer, so it won't have to be relinked.
-	$(MAKE) CC="stage1/xgcc -Bstage1/" CFLAGS="$(BOOT_CFLAGS)" LDFLAGS="$(BOOT_LDFLAGS)" libdir=$(libdir) STAGE_PREFIX=stage1/ LANGUAGES="$(LANGUAGES)"
+	if test $@ = bootstrap-ultralean; then \
+		languages="$(BOOT_LANGUAGES)"; else \
+		languages="$(LANGUAGES)"; fi; \
+	$(MAKE) CC="stage1/xgcc -Bstage1/" CFLAGS="$(BOOT_CFLAGS)" LDFLAGS="$(BOOT_LDFLAGS)" libdir=$(libdir) STAGE_PREFIX=stage1/ LANGUAGES="$$languages"
 	$(MAKE) stage2
-	-if test $@ = bootstrap-lean; then rm -rf stage1; else true; fi
+	-if test $@ = bootstrap; then true; else rm -rf stage1; fi
 	$(MAKE) CC="stage2/xgcc -Bstage2/" CFLAGS="$(BOOT_CFLAGS)" LDFLAGS="$(BOOT_LDFLAGS)" libdir=$(libdir) STAGE_PREFIX=stage2/ LANGUAGES="$(LANGUAGES)"
 
 bootstrap2: force


Thu Jan 15 17:37:23 1998  H.J. Lu  (hjl@gnu.org)

	* Makefile.in (bootstrap-ultralean): New target.

Index: Makefile.in
===================================================================
RCS file: /home/work/cvs/gnu/egcs/Makefile.in,v
retrieving revision 1.1.1.8
diff -u -r1.1.1.8 Makefile.in
--- Makefile.in	1997/12/08 17:03:45	1.1.1.8
+++ Makefile.in	1998/01/16 01:43:34
@@ -1290,8 +1290,8 @@
 # In theory, on an SMP all those dependencies can be resolved
 # in parallel.
 #
-.PHONY: bootstrap bootstrap-lean bootstrap2 bootstrap3 bootstrap4
-bootstrap bootstrap-lean bootstrap2 bootstrap3 bootstrap4: all-texinfo all-bison all-byacc all-binutils all-gas all-ld
+.PHONY: bootstrap bootstrap-lean bootstrap-ultralean bootstrap2 bootstrap3 bootstrap4
+bootstrap bootstrap-lean bootstrap-ultralean bootstrap2 bootstrap3 bootstrap4: all-texinfo all-bison all-byacc all-binutils all-gas all-ld
 	@r=`pwd`; export r; \
 	s=`cd $(srcdir); pwd`; export s; \
 	$(SET_LIB_PATH) \
@@ -1302,7 +1302,7 @@
 	case "$@" in \
 	  *bootstrap4 ) msg="Comparing stage3 and stage4 of the compiler"; \
 	  		compare=compare3 ;; \
-	  *-lean )	msg="Comparing stage2 and stage3 of the compiler"; \
+	  *-lean | *-ultralean ) msg="Comparing stage2 and stage3 of the compiler"; \
 	  		compare=compare-lean ;; \
 	  * )		msg="Comparing stage2 and stage3 of the compiler"; \
 	  		compare=compare ;; \

Re: A bootstrap-ultralean patch

[Joe Buck]

> I made this patch to allow
> 
> # make bootstrap-ultralean
> 
> which only builds the C compiler for stage 1. It saves both CPU
> time and disk space.

The problem with this is that the resulting binary will differ from
the binary produced by a normal bootstrap.  It would be particularly
troublesome if you (or anyone else) were ever to distribute a
binary release of a gcc produced by this method, resulting possibly
in bug reports that cannot be duplicated.

At least do stage 2 (the only purpose for stage 3 is the check).

Re: A bootstrap-ultralean patch

[H.J. Lu]

> 
> > I made this patch to allow
> > 
> > # make bootstrap-ultralean
> > 
> > which only builds the C compiler for stage 1. It saves both CPU
> > time and disk space.
> 
> The problem with this is that the resulting binary will differ from

I think "will" is a strong word. ultralean still compares the
C part. Someone like me is willing to take the risk of not
comparing the C++, Objc and Fortran parts.

> the binary produced by a normal bootstrap.  It would be particularly
> troublesome if you (or anyone else) were ever to distribute a
> binary release of a gcc produced by this method, resulting possibly
> in bug reports that cannot be duplicated.
> 
> At least do stage 2 (the only purpose for stage 3 is the check).


-- 
H.J. Lu (hjl@gnu.org)

Re: A bootstrap-ultralean patch

[Jeffrey A Law]

  In message < m0xt18j-0004ecC@ocean.lucon.org >you write:
  > Hi,
  > 
  > I made this patch to allow
  > 
  > # make bootstrap-ultralean
  > 
  > which only builds the C compiler for stage 1. It saves both CPU
  > time and disk space.
Errr, the existing bootstrap targets should only be building C during
stage1.  If that's not working we need to fix them instead of creating
a new target.
jeff

Re: A bootstrap-ultralean patch

[H.J. Lu]

> 
> 
>   In message < m0xt18j-0004ecC@ocean.lucon.org >you write:
>   > Hi,
>   > 
>   > I made this patch to allow
>   > 
>   > # make bootstrap-ultralean
>   > 
>   > which only builds the C compiler for stage 1. It saves both CPU
>   > time and disk space.
> Errr, the existing bootstrap targets should only be building C during
> stage1.  If that's not working we need to fix them instead of creating
> a new target.

Ooops. I meant to say stage 2.

-- 
H.J. Lu (hjl@gnu.org)

Re: A bootstrap-ultralean patch

[Jeffrey A Law]

  In message < m0xu2GT-0004ecC@ocean.lucon.org >you write:
  > > Errr, the existing bootstrap targets should only be building C during
  > > stage1.  If that's not working we need to fix them instead of creating
  > > a new target.
  > 
  > Ooops. I meant to say stage 2.
In that case I could probably recommend against the patch.

Jeff

Re: A bootstrap-ultralean patch

[Jeffrey A Law]

  In message < m0xu2GT-0004ecC@ocean.lucon.org >you write:
  > > Errr, the existing bootstrap targets should only be building C during
  > > stage1.  If that's not working we need to fix them instead of creating
  > > a new target.
  > 
  > Ooops. I meant to say stage 2.
In that case I could probably recommend against the patch.

Jeff